ACKNOWLEDGMENT during Catalytic Oxidation Reactions over a Supported

Platinum Catalyst, A.C.S. Symp. Ser., 65, 475 ( 1978).

Financial support was provided by National Science Founda-
tion Grant No. ENG-75-19475 and The University of Connecti-
cut Research Foundation. Some of the experimental measnre-
ments were conducted in the Department of Chemical Engi-
neering at Cambridge University, Cambridge, England, with
the cooperation of Dr. C. N. Kenney. Catalyst properties were
measured in the laboratory of Prof. J. B. Butt, Northwestern
University, Evanston, Illinois.
LITERATURE CITED
Al-Taie, A. S., and L. S. Kershenbaum, Effect of Periodic Op-
eration on the Selectivity of Catalytic Reactions, A.C.S.
Symposium Ser., 65,512 ( 1978).
Bailey, J. E., Periodic Operation of Chemical Reactors: A
Review, Chem. Eng. Commun., 1, 111 (1973).
-, Periodic Phenomena, in Chemical Reactor Theor!/-
A Reoiew, L. Lapidus and N. R. Amundson, ed., p. 758,
Prentice-Hall, Englewood Cliffs, N.J. ( 1977).
Bilimoria, M. R., and J. E. Bailey, Dynamic Studies of Acetyl-
ene Hydrogenation on Nickel Catalysts, A.C.S. Symp. Ser.,
65,526 ( 1978).
Carberry, J. J., Chemical and Catalytic Reaction Engineering,
McGraw-Hill, New York ( 1976).
Cutlip, M. B., C. J. Hawkins, and C. N. Kenney, Elementary
Step Models and Multiplicity in Surface Rate Process, paper
presented at A.1 Ch.E. Chicago meeting, 111. ( Dec., 1976).
Cutlip, M. B., and C. N. Kenney, Limit Cycle Phenomena
Douglas, J. M., Process Dynamics and Control, VoI. 2, Prentice-
Hall, Englewood Cliffs, N. J. (1972).
Hanson, F. V. and J. E. Benson, An Inexpensive Noncon-
taminating Gas Recirculation Pump, J. Catalysis, 31, 471
(1973).
Oh, S. H., et al., Carbon h4onoxide Oxidation in an Integral
Reactor: Transient Response to Concentration Pulses in the
Regime of Isothermal Multiplicities, A.C.S. Symp. Ser., 65,
461 ( 1978).
Renken, A,, H. Helmrich, and K. Schugerl, Beeinflussung der
eff ektiven Geschwindigkeit heterogen-katalytischerReaktionen
durch aufgezwungene Konzentrationsschwankungen, Chem.
Ingr. T e c h . , 46,647 (1974).
Renken, A,, M. MuIIer, and C. Wandrey, ExperimentaI Stud-
ies on the Improvement of Fixed-Bed Reactor by Periodic
Operation-the Catalytic Oxidation of Ethylene, Proc. of
the 4th IntemationaVGth European Symp. on Chemical Re-
action Engineering, Dechema, Frankfurt ( 1976).
Sheintuch, M., and R. A. Schmitz, Oscillations in Catalytic
Reactions, Cat. Reu. Sci. Eng., 15, 107 (1977).
Unni, M. P., R. R. Hudgins, and P. L. Silveston, Influence of
Cycling on the Rate of Oxidation of SO2 Over a Vanadia
Catalyst, Can. J. Chem. Eng., 51, 623 (1973).
Wei, J., The Catalytic Muffler, Adsances in Chemistry Series,
148, 1 ( 1975).
Manuscript receitied June 2, 1978; revision feceiued Decenther 19,
and accepted February 6 , 1979.

Mixing with Helical Ribbon Agitators

Of necessity, the mixing process is sometimes restricted to the laminar W. IAN PATTERSON
regime, although turbulent mixing is generally more desirable. Common
examples of laminar mixing are found when the fluid has a very high vis- P. J. CARREAU
cosity, or when one of the mixture components is shear sensitive. I t has been and
pointed out that the helical ribbon agitator (HRA) is admirably suited to the
low Reynolds number mixing process. This work derives a model to predict C. Y. YAP
the power consumption of the HRA. The model has been developed with Department of Chemical Engineering
the aid of experimental data and tested extensively using literature data. For Ecole Polytechnique of Montreol
a wide range of mixer geometries and sizes, it predicts power consumption Montreal, Canoda
with an average deviation of 137&
The concept of relative efficiency of mixers is also described as an aid to
comparing different HRA mixer geometries. Finally, the problem of scale-up
of different HRA configurations is discussed.

Part II. Newtonian Fluids

SCOPE
The mixing of very viscous fluids is qualitatively and
quantitatively quite different from the process of blending
low viscosity fluids. Parker (1964) noted that the helical
ribbon agitator (HRA) is particularly well suited for mix-
ing viscous liquids where the mixer flow is laminar. This
observation is supported by the data of Gray (1963) and
Correspondence concerning this paper should he addressed to Pierre
Carreau. C. Y. Yap is with the University of Mnlaysi,?.
Part I of this paper was published in Carl. I. Chcm. Eng., 54, 135
(1976).
62 The American Institute of Chem-
0001-1541-79-2552-0508-$01.05.
ical Engineers, 1979.
Johnson (1967). The choice of a mixer configuration is
governed by three characteristic parameters: mixing time,
power consumption, and mixer efficiency. The efficiency
of a mixer is related to the total energy required to achieve
a given degree of homogeneity. Mixing efficiency has been
virtually ignored by previous workers, undoubtedly due
to the impossibility of finding a usable definition of the
(thermodynamically) perfect mixing process. This diffi-
culty can be avoided by defining an efficiency relative to
an arbitrary, standard mixer. This work proposes a stan-
dard mixer. This work proposes a standard mixer, and the
relative efficiency (eff),,, has been used to evaluate five
HRAs of different geometries.

Page 508 May, 7979 AlChE Journal (Vol. 25, No. 3)

 Models for predicting the power consumption of HRAs
have been proposed by Nagata et al. (1957), Bourne and
Butler (1965, 1969), Chavan and Ulbrecht (1973b), and
Hoogendoorn and Den Hartog (1967). [The state of mix-
ing science has been exhaustively examined in Nagatas
(1975) monograph.] These models are not generalized in
the sense that one or more geometrical or fluid parameters
have been neglected. This is not surprising, since dimen-
sional analysis shows that ten dimensionless groups must
be considered in the complete analysis. This paper pro-
poses a general model developed from a drag flow analysis
about the blades, using the observed hydrodynamics of
CONCLUSIONS AND SIGNIFICANCE
The problem of defining and measuring the absolute
efficiency of the mixing process has been avoided. A rela-
tive efficiency based on an arbitrary reference mixer was
defined and has been found useful for comparing the
performance of different mixers. In this way, it was found
that the impeller efficiency increased as the blade width
increased and the blade-vessel wall gap increased. A
single-blade impeller was found to be significantly (-2.5
times) more efficient than the twin-blade agitator of similar
geometry.
A model for the power consumption of the helical ribbon
agitator was developed from drag flow analysis and the
observed hydrodynamics of the flow. The original mode!
could be drastically simplified by taking average values
of some of the parameters that have little effect on the
power consumption. The resulting model gave excellent
agreement for the eight different mixer geometries and
the three fluids examined. This model was slightly modified
so as to be comparable to results reported in the literature;
then the average deviation between experimental and
predicted values was found to be 13% for fourteen differ-
ent geometries. Part of this deviation may be ascribed to
error in the approximate calculation of the blade length
Mixing is a process that is basic to many industries
and has been the subject of much investigation. Holland
and Chapman ( 1966), Oldshue ( 1966), Rushton (1954),
and Voncken ( 1965) have each compiled comprehensive
reviews of fluid mixing, and Bourne (1964) has reviewed
the mixing of powders, pastes, and non-Newtonian fluids.
With the exception of Bournes review, most of the pub-
lications that have appeared are concerned with the mix-
ing of Newtonian fluids and have primarily discussed
turbulent mixing. This is not surprising, since turbulent
mixing is a much quicker process than mixing done in
the laminar regime. Nonetheless, there are situations
where mixing is restricted to the laminar regime. Ex-
amples of this situation include cases where the fluid
has a very high viscosity, such as is possessed by polymer
solutions or melts, or where one or more of the mixture
components is shear sensitive. These cases are character-
ized by a low Reynolds number diiring mixing, where
mixing proceeds through fluid deformation and not by
turbulent eddy diffusion.
Parker (1964) in an article on modern practice on
mixing, reported the recent introduction of the helical
ribbon agitator which was effective in overcominq the
large viscous forces present in the mixing of high vis-
AlChE Journal (Vol. 25, No. 3)
the fluid flow during mixing. Most practical HRAs have
certain dimensionless groups that vary only within narrow
limits; thus, for engineering purposes, it is possible to
obtain a simplified model employing only three dimension-
less groups. This model has been verified with eight differ-
ent geometries and three fluids. A slightly altered model
was checked for fourteen different mixing conditions ob-
tained from literature.
Finally, since almost all work done on HRA mixers has
been on a laboratory scale (including this work), the ques-
tion of scale-up is briefly examined.
to diameter ratio l / d not reported by the previous workers.
The most important parameters affecting the power
consumption were found to be the diameter ratio of the
agitator to vessel d / D , the blade length to diameter ratio
l / d , and the agitator pitch. Power consumption increased
as the gap between the agitator and vessel wall decreased
(increased d / D ) and as the blade length increased (increas-
ing V d ) . Conversely, increasing the pitch ratio of the agi-
tator decreased the power consumed, but this was at the
expense of the mixing efficiency. Interestingly, the power
consumption was not strongly affected by the blade width.
However, the pumping capacity of the wide bladed agi-
tator was markedly superior, giving this impeller a high
efficiency.
The data and results reported here apply to a left-handed
helical agitator rotated anticlockwise; that is, the impeller
pumping action was downwards at the vessel wall. A few
data were taken with the same impellers rotated in the
opposite direction, pumping upwards at wall, and it was
found that the power consumed was approximately 10%
greater for this condition. This effect is not accounted for
in the proposed model which was characterized for the
downwards pumping flow only.
cosity fluids. The pertinent geometry of this type of
agitator is shown in Figure 1.
As shown by many authors (Coyle et al., 1970; Gray,
1963; Hoogendoorii and Den Hartog, 1967; h4oo-Young
et al., 1972; Nagata et al., 1957, 1971), the time necessary
to achieve a certain degree of homogeneity (mixing time)
in viscous fluids is much shorter with a helical ribbon
impeller than with a turbine (most commonly used)
or with other impellers. The helical ribbon agitator and
the helical screw draught tube system (Chavan and
Ulbrecht, 19734) are probably the most efficient mixers
in the laminar regime. The inclined surface of the ribbon
moving near the vessel wall produces axial movement of
the liquid as well as tangential flow (Gray, 1963). Thus,
a relatively effective top to bottom flow motion is ob-
tained, compared with other impeller types.
Although the power requirement of the helical ribbon
mixer is much higher than that of turbines or propellers
at the same Reynolds number (Gray, 1963; Hoogendoorn
and Den Hartog, 1967; Johnson, 1967), the energy re-
quired to obtain a given degree of homogeneity is con-
siderably less. Gray (1963) compared mixing and power
data for various impellers and concluded that the helical
ribbon agitator mixes the most rapidly and requires the
May, 1979 Page 509

I a
SP
:ig. 1. Sketch of helical ribbon agitator system.
least energy. This was confirmed by Johnson (1967)
who also showed that the helical ribbon mixers have a
definite advantage for very viscous fluids. His data, ob-
tained with corn syrup of 2 500 centipoises, indicated
that at the same mixing effectiveness (mixing time of 5
min), the power required for the helical ribbon agitator
was one surth to one twelfth of the power required for
turbines and propellers.
The mixing time requirement has been determined for
a number of helical ribbon agitator geometries (Carreau
et al., 1976), and this presentation is restricted to the
development of a model to predict the power consumption
of helical ribbon agitators mixing Kewtonian fluids. The
model has been tested using values reported in the
literature in addition to data measured by us. A short
discussion on the relative efficiencies of different helical
ribbon agitators concludes the presentation.
THEORETICAL CONSIDERATIONS
Although we are primarily concerned with the predic-
tion of power consumption, the test of a mixer's utility
is its efficiency; that is, how much energy must be ex-
pended to obtain a specified degree of homogeneity.
Thus, we digress temporarily to consider mixing time
and relative mixing efficiency.
Mixing Time
The mixing time t, is usually defined as the time re-
quired to reach a specified level of uniformity in a
mixing system. It is a quantitative measurement of the
time necessary to reduce the intensity of segregation
(Danckwerts, 1953; Lacey, 1954) to a speciiied value
which is usually taken in the order of 0.1 to 1%.
For nonturbulent mixing, however, the intensity of
segregation is difficult to measure because of the dif-
ficulties involved in sampling. Thus, various other criteria
for mixing time have been developed. The methods for
measuring mixing time include electrical conductivity
changes, refractive index differences, color differences,
indcator changes during an ionic reaction, and tempera-
Page 510 May, 1979
ture differences (Ford et aI., 1972; Voncken, 1965; Hoo-
I
gendoorn and Den Hartog, 1967). Nevertheless, because
of the different criteria used with different methods, the
effects of additives, and differing sampling philosophies,
1
the mixing times obtained cannot be taken as absolute
values. They are useful primarily as relative values, de-
fined for a particular system and experimental technique.
Relative Mixing Efficiency
d The mixing efficiency is an important parameter of an
agitation system. Of course, the decisive factor in the
evaluation of the efficiency of a mixer is the degree of
homogeneity obtained after a certain time. The mixing
efficiency of a particular mixer is relatively high when
a minimal power and a short mixing time are required
to achieve a specified degree of mixing. In some instances,
the mixing efficiency can be extremely low, for example,
when whirling and fluid motion are produced only
locally without contributing in any way to the mixing
within the main body of the fluid.
Since it is impossible to define a perfect mixer, a rela-
tive mixing efhciency is deSned with respect to the
performance of an arbitrary, hypothetical mixer.
In a mixing vessel, the power consumption for laminar
flow is given by
P = C,pN2d3 (1)
and the mixing time t, for helical ribbon agitators is
inversely proportional to the rotational speed N of the
impeller (Coyle et al., 1970; Chavan and Ulbrecht, 1973;
Gray, 1963, Hoogendoorn and Den Hartog, 1967; John-
son, 1967; Moo-Young et al., 1972; Nagata et al., 1957) :
"
Therefore, the energy input in the vessel is
We define a hypothetical reference mixer by arbitrarily
choosing the product CICz equal to 1000. This value is
chosen from literature data as the lower estimated limit
of the product of the power consumption and mixing time.
The efficiency of a given system when compared to
our reference mixer is the relative efficiency defined by
where the product Pt,, is the actual energy required to
obtain a given degree of mixing.
Development of o Power Consumption Model
The model is based on the principle that the torque
exerted on a rotating impeller is due to the drag force
exerted by the fluid flow around the impeller blade. This
is in contrast to the models proposed by Bourne and
Butler (1965) and Chavan and Ulbrecht ( 1973b), where
it is assumed that the power consumed is that resulting
from a Couette flow, that is, flow between two coaxial
cylinders,
The observed flow patterns that form the basis of
the model are illustrated in Figure 2. The horizontal
motion of the fluid is qualitatively illustrated in F'g tires
11
2a and 20 for fixed coordinates and relatively to the
blades, respectively. The blade velocity is much larger
than the fluid velocity.
The forces exerted by tlie fluid on the blade at a given
point are caused by the dynamic pressure of the stream
veIocity whose action is normal to the blade (the resulting
force is generally defined as normal drag or form drag)
AlChE Journal (Vol. 25, No. 31
and the tangential stresses which result in the friction FLUID FLUD RELATIVE
drag or skin drag as the fluid element slides along the
surface of the blade and which finally departs owing to
its relative motion with respect to the blade. It is evident
that the total torque exerted on the impeller is the product
of the impeller radius and the sum of the two drag forces
over the total impeller blade surface. The torque exerted
on the impeller blades is given by

T = JJA - In.(: + p8)lordA

- (5)
+
The term - [n*(z p S ) l e is the component of the a1 FIXED COORDINATES b) RELATIVE COORDINATES
force per unit area i n the 7 direction, which is the direc- Fig. 2. Tangentiol velocity profiles.
tion of impeller rotation. However, since r does not vary
must for helical ribbon blades, Equation (5) is approxi- Reb is taken to be
mated by
r = Fk& (6) P
and the power is given by and the range covered corresponds to the values of our
experimental data in the laminar region where 5 < Reb
P = 7 (27rN) = %rRbNFg, (7) < 25. For tangential flow, the assumptions imply that,
on the average, the fluid flows tangentially over the
The drag force can be expressed in terms of the drag blades on a distance equal to w and covers a width equal
coefficient: to w / 2 . From qualitative visual studies of the flow about
the blades, these assumptions appear to be reasonable.
From Equations (7), ( l o ) , ( l l ) , and (12), it fol-
lows that the total power is given by
Unfortunately, it appears that no data or correlations
on the drag flow about an inclined plane are available P = 4nbp#NRb (N,R,) 2h[
CD, sin Jr + % CD,cos $1
in the literature. To obtain an approximate expression
for the e component of FI,, it is assumed the total drag (16)
is the sum of two drag forces, one for a plate normal Assuming that R, N Rb [an assumption already implied
to the main flow field and the second for a plate parallel in Equation ( S ) ] , we obtain the following expression
to the How. for the power number:
It is further assumed that the approaching fluid velocity
is in the B axis [in Carreau et al. (1976) it was shown
that the main component of the How was tangential];
that is
V, N 27rN,Rc = 2n( N - N j ) Rc (9)
It follows that
In order to calculate the power or the power number, one
F K ~ 2p (flN&) [CD,& -b C ~ t A t l (10) needs the value of N , the relative rotational velocity of
The characteristic area of the blades for the normal flow the fluid, which can be obtained from experimental data.
is appropriately defined by However, in general, the fluid velocity is much less than
that of the impeller, and as a first approximation, we
An = nb(lsin$)w (11) can set N , equal to N . (This is especially true at low
values of N . ) Thus, the power can be calculated from
For the tangential flow, the characteristic area is not the fluid properties, dimensions of the mixing system,
as well defined, since an incoming fluid element slides and impeller rotational speed. This approximation would
over the blades in the radial and tangential directions predict a too large power consumption for high impeller
at the same time. We take rotational speeds. A suitable expression for N , can be
developed with the aid of some experimental data.
At = nb(Z cos $) -
W
2
(12)
MOTOR
assuming that, on the average, the flow is divided in two
halves on the front side of the blades and that very little GEAR \ c w SPEED CONTROLLER

drag results from skin friction on the backside of the TEMPERATURE DIGITAL FREQUENCY
METER
inclined blades. CONTROLLER
TORQUE TRANSDUCER
The expressions for the drag coefficient are obtained TORQUE RECORDER
from Whitaker (1968) :
For normal flow

For tangential flow

In both cases, the Reynolds number around the blades Fig. 3. Experiment01 opporotus.

AlChE Journal (Vol. 25, No. 3) May, 1979 Page 511

 TABLE
1.IMPELLERS
CHARACTER~STICS

A I 0.130 0.145 0.137 6.35 1.11 4.48 0.097 0.719

B I1 0.130 0.145 0.137 6.35 1.11 3.00 0.097 1.048
C I11 0.130 0.145 0.137 6.35 1.11 4.12 0.195 0.707
D IV 0.105 0.145 0.137 6.35 1.37 4.00 0.121 0.848
E V 0.130 0.145 0.137 6.35 1.11 4.39 0.097 0.695
F VI 0.222 0.248 0.234 9.53 1.11 4.44 0.099 0.690
G VIII 0.219 0.248 0.238 9.53 1.12 4.75 0.072 0.724
H VI 0.222 0.291 0.234 9.53 1.30 4.44 0.099 0.690

TABLE
2.FLUIDS
PROPERTIES and impeller IV has a smaller outside diameter. Impeller V
is a single-blade agitator with all dimensions identical to
P a impeller I. Geometry F with impeller VI in the 0.248 m
Fluid ( kg/m3 1 ( N s/m2) vessel is a scaled-up version of geometry A. Geometry G
also uses the 0.248 m vessel, but the agitator has a slightly
100% glycerol 1 254 0.568 larger diameter and a smaller blade width compared to com-
100% glycerol 1254 0.708 bination F. Geometry H is the combination of impeller VI
100% glycerol 1259 0.800 with the larger vessel of 0.291 m diameter.
Silicone oil 1100 0.137 ;A tabulation of the fluids used is given in Table 2.
Vitrea oil 869 0.193
RESULTS AND DISCUSSION

EXPERIMENTAL The power consumption model requires a determina-

tion 01 the relative nuid-agitator velocities to be use-
The apparatus employed in this study is shown schematically ful. The dimensionless analysis of mixing of White and
in Figure 3. It consisted of a water bath with the mixing Hrenner ( 1934) suggests the following correlation when
vessel immersed in it. The fluid to be mixed was placed in the Froude number may be neglected in laminar mixing:
the cylindrical vessel and the motor-agitator assembly lowered
as a unit to place the agitator in the vessel. Three cylindrical
mixing vessels were used in this study. The smallest had
dimensions of 0.145 m internal diameter and a height of
5= a"[ I
N
I
d2N p ]"'[ ["I: fl""[:Ia4 (18)

0.203 m, and the second was 0.248 m internal diameter and The five indexes of Equation (18) were determined
0.305 m high. The largest vessel of 0.291 m internal diameter
had a height of 0.38m. Ly a multiple linear regression of the experimental data
The agitator was driven by a reversible, variable speed of the tangential velocity profiles (see Carreau et al.,
motor coupled to the agitator through a right-angle speed 1970) to give
reducer to give minimum and maximum agitator speeds of
Nr- - 1
0 and 500 rev/min, respectively. The rotational speed was -
[ IcL(;( ( ; 4)
detected by a magnetic transducer, and the torque acting N
on the agitator was measured by a shaft mounted dynamic d2Np 14 10.6 10.176
torquemeter. The reader is referred to Yap (1976) for a
detailed description of the apparatus and methods, (19)
Eight helical ribbon ag'tators constructed from stainless and the fit is shown in Figure 4. The correlation coef-
steel were used in this study. They were made by rigidly fixing
one or two stainless steel helices (the blades, see Figure 1)
t o the shaft. In the case of the twin-blade agitators, the
helices were mounted 180 deg apart. The details of the im-
pellers' geometries are given in Table 1. Geometries A to E
are the combinations of impellers I to V, respectively, with

1
the same 0.145 m vessel. Impeller 11 has a larger pitch com-
pared to impeller I, while impeller I11 is wider in blade width

3 , 1 . , , , , GEOMETRY
NP 10' -

A A GLYCEROL 0 5 6 8 N 5 m-'

2
0
D
A
B
c
D
'8

"
''
' I
.
E

~
! F
H
"

"

'I
0,

M,N,

05

-20 ,
101 I I I I , I 1 1 1 1 I I ' I I I I I I I I I l i l l l

3 5 10 20 50 100 100 101 102 103

d 2 N P ) 0 9 3 ~ ~ ) 0 9d1 I 2 3
(-
(I) P
Fig. 5. Generalized correlation of the power number
dictions of Equation (21).
- pre-

Page 512 May, 1979 AlChE Journal (Vof. 25, No. 3)

ficient is 0.98, with a standard deviation of 0.01, These
values are quite acceptable when we consider the scatter
of the experimental data. This correlation is obviously
restricted to Newtonian fluids and the range of the
experimental variables.
Equation (17) proposed for the power consumption
can now be solved by substituting the above correlation I
a
of the fluid velocity to obtain e Ov)
m t-
L? mw
+ 1.8 cos $3
Np = 7'9 " ( 3 ) [(
0.16

d2: )O.M ; (-
[5.8 (Reb)0.Z6 sin t,b
(- )ll>Jl fi)1.23]

(20)
This equation can be simplified on the following grounds:
1. For commonly used helical ribbon agitators, w / d N
0.1, and t,b N 15 deg. Therefore, we set (w/d)OJ6 N
0.69,sin $ 1:0.258, and cos t,b N 0.965.
2. Within the range of the experimental conditions,
Reb0,26varies from 1.32 to 2.17. Hence, we take Reb0.2s
N 1.82 as an average value.
Equation (20) then reduces to

Np = 24 nb [ ( d2Np
-)0.93 (3""
(T ) ] d 1.23 -1

Although this result is less accurate than Equation

(20), it is much more attractive because of its simplicity.
It contains the two most important geometrical ratios:
D / d , which expresses the influence of the vessel wall-
impeller blade gap, and d/Z, which characterizes the
geometry of the impeller. The influence of the pitch
or the inclination angle of the blades is taken into account
through the ratio d/Z. The blade width has only a small
influence on the power consumption, as indicated by
the exponent 0.16 of the ratio w / d . This is surprising,
but it can be explained as follows. Although the total
blade area submitted to drag action varies linearly with
w , the drag coefficients CD,and C D , drop as to increases,
as shown by Equations (13) and (14). Moreover, the
fluid tangential velocity increases slightly with w as
shown by the correlation, Equation (19). Thus, the net
effect of w on the power consumption is not very im-
portant.
The usefulness of Equation (21) was tested using all
the power consumption data of the eight geometries with
glycerol, silicone, and Vitrea oil, The result is shown in
Figure 5; Equation (21), which is the solid line, cor-
relates the data very well with a standard deviation of
only 6%. The data for the one-blade impeller (geometry
,E) have been multiplied by 2, so they could be correlated
by the same expression.
Comparison with Literature Data
It is interesting to note that in the power correlation
[Equation (21) ] the exponent of the Reynolds number
is slightly greater than the value of -1 reported by most
authors. The value of -0.93, although only 7% different
from -1, considerably improves the fit of our data.
Correlations of power consumption proposed in the
literature to date have been of the form
Np = KRe-1 (22)
and, in order to test the validity of the present model
for the helical ribbon systems used by previous researchers,
we force Equation (21) to take the form of Equation
(22) ; that is

AlChE Journal (Vol. 25, No. 3) May, 1979 Page 513

 4.RELATIVEEFFICIENCY
TABLE OF VARIOUS IMPELLERS
(Glycerol, p = 0.568 N * s/mz)
Impeller nb D/d l/d w/d
I 2 1.11 4.48 0.097
I1 2 1.11 3.00 0.097
I11 2 1.11 4.12 0.195
IV 2 1.37 4.00 0.121
V 1 1.11 4.39 0.097
0 Values obtained from Carreau et al. (1978).
By curve fitting the data of Figure 5, the value of C3
was found to be equal to 60 for twin-blade impellers;
A comparison of the values of K predicted by Equa-
tion ( 2 4 ) with the experimental values reported in the
literature for a variety of helical ribbon agitators with
Newtonian fluids is presented in Table 3. Generally, the
percentage deviation between the predicted and experi-
mental values is very acceptable when we consider the
experimental errors and the imprecision of 1, the length
of the blades. It was necessary to calculate 1 from the
pitch and the diameter of the blades, since the factor Z/d
is not reported in the literature. Nevertheless, almost
half of the predicted values are within 2 8 % of the
experimental values. It should be noted that the larger
deviations are observed for geometries HG-1, HG-2,
RB-1, and 2-1 which had the impeller pitches consider-
ably different from those commonly used. The correla-
tion could be improved for these geometries by correcting
for the inclination angle. Table 3 highlights the observa-
tion that geometry has a considerable inHuence on the
power consumption. An example is geometry Z-1 which
has a value of K eight times larger than that of geometries
HG-4 and RB-1. The constant K , however, is independent
of the fluid properties lor Newtonian fluids.
A few tests were made with the agitator rotating in
the clockwise direction (impeller pumping upwards at
the vessel wall). The power consumed was found to be
about 10% higher than the identical situation with
counterclockwise rotation. It is believed that depending
on the direction of the agitator rotation, the inertia forces
affect differently the flow patterns; hence, Equation (19)
should be slightly modiiied to account for that effect.
Relative Efficiency of Vorious Agitators
From the known mixing time (Carreau et al., 1976)
and the power consumption correlation, we can estimate
and compare the relative efficiency oh each system in the
0.154 m vessel.
The concept of relative efficiency defined above is
used for the calculation. From Equations ( 2 ) , (4), and
(24)
The relative efficiencies of the five impellers for glycerol
are tabulated in Table 4.
Page 514 May, 1979
(Effi-
C2* ciency),,]
45 0.055
49 0.100
22 0.137
53 0.086
58 0.094
0.1 I , I I I l l I I 1
0.2 0.5 I 2 5
N, SECONDS-'
Fig. 6. Influence of geometry on power consumption for glycerol
( p = 0.568 N.s/mZ).
The higher the relative efficiency, the less total energy
is required to achieve a certain degree of mixing in a
given volume of fluid. Table 4 indicates that impeller
111, which has wider blades ( w is twice of that of the
others), is the most efficient impeller; its efficiency is 2.5
times that of impeller I. A larger blade width results
in a higher pumping capacity, but, as discussed earlier,
the torque is unchanged. The relative efficiency of im-
peller I1 is rather good as it consumes low power, since
the ratio Z/d is rather small (high pitch of the blades)
compared with the others. The efficiency of impeller Iv
is better than that of impeller I; because of the larger
gap between the vessel wall and the blades (23% larger),
the power consumed is considerably less. The one-blade
impeller V has only a slightly lower pumping capacity
or lower mixing effectiveness; however, since its power
consumption is approximately half of that consumed by
impeller I, its efficiency is almost doubled.
Scale-up a n d Design Aspects
Since it is neither convenient nor economical to ex-
periment with industrial size units to get optimal process
conditions, it is desirable to have a scale-up technique
to predict the performance of an industrial scale mixer
from experiments carried out on laboratory size equip-
ment.
In this section, the scale-up possibility of the systems
studied is discussed. The implication and extrapolation
of the variables studied and their effects are briefly re-
viewed. To facilitate the discussion that follows, power
consumption data are plotted in a noncorrelated form in
Figure 6, showing the effects of the geometrical variables
for glycerol.
To predict the scale-up performance of agitators in
production size mixing systems, information about the
quality or degree of mixing and energy consumed should
be compared. Dimensionless correlations available can
be used to conveniently scale-up geometrically similar
systems. In the so-called laminar mixing with helical
AlChE Journal (Vol. 25, No. 3)
ribbon agitators, Nt,, which is the number of impeller wider blades may appear interesting from the point of
revolutions needed to obtain an arbitrary but constant view of efficiency, they are difficult to build and hence
degree of homogeneity, is constant for a given geometry expensive.
[Equation ( 2 ) 1.
Thus, for geometrically similar systems, the impeller Effect of the Number of Blades
rotational speed to maintain a given mixing time is inde- For the single-blade helical ribbon impeller, the mixing
pendent of the scale of the system. The power number is time is about 20% longer than with the similar twin-
inversely proportional to the Reynolds number, as ex- blade impeller, at the same rotational speed. However,
pressed by Equations (22), (23), or (24). The com- the power consumed for the one-blade impeller is approx-
bination of Equations ( 2 ) and (24) gives imately half of that consumed by the two-blade impeller.
Its efficiency is almost double, and hence one-blade im-
pellers could be quite interesting for design purposes.
More experimentation on one-blade impellers is required
in order to draw definite conclusions.
where for Newtonian fluids Cz and K are functions only
of the geometrical ratios. For a given mixer geometry, ACKNOWLEDGMENT
d / D = C3, a constant; hence
The authors wish to acknowledge the financial assistance pro-
tm2P vided by the Canadian National Research Council (grant A-
-= C2C3K = constant 5817). They also wish to thank Messrs J. Coderre, J. MorBno,
D3P and B. Chartrand for their most appreciated technical help.
Therefore, the power per unit volume required to ob-
tain a certain homogeneity is independent of the vessel NOTATION
size and is proportional to the fluid viscosity. UO,~ 1 UZ,
, u3, u4 = indexes in Equation (18), dimensionless
Effects of the Geometrical Rotio D/d and l a n k Dlameter A = impeller blade area, m2
The effect of the gap between the vessel wall and the A, = impeller blade area normal to the flow, m2
impeller blades is accounted for by the geometrical ratio At = impeller blade area tangential to the flow, m2
ClCz and C3 = proportionality constants, dimensionless
D / d . The effect of D / d in the range 1.11 D / d 4 1.37
(this work) is predicted accurately by the proposed equa- CD = total drag coefficient, dimensionless
tion. The predictions for values of 1.01 A D / d 6 1.11 CD, = normal drag coefficient, dimensionless
(literature data, Table 3 ) vary from good to accurate. CD, = tangential drag coefficient, dimensionless
Mixing systems with a scale-up factor of 5 (that is, 0.76 m
d = diameter of impeller, m
diameter vessels) can well be described by the results d, = diameter of impeller shaft, mm
of this work. The effectiveness obtained for our 0.154 m
D = diameter of vessel, m
and 0.254 m vessels is comparable to that obtained by
FK = drag force, N
Coyle et al. (1970) in a 0.76 m tank, with D / d N 1.03; F K ~= drag force in the 0 direction, N
the power predicted by Equation (24) for the 0.559 m
h = height of impeller, m
vessel of Hall and Godfrey (1970) is within 2.4% of the
H = height of liquid in vessel, m
data. The performance of large scale mixing systems is
K = proportionality constant defined by Equation
(22), dimensionless
predictable for geometrical ratios in the range of the 1 = length of impeller blade, m
values studied in this work. n = unit vector normal to the blade surface
Effect o f the Pitch Rotio nb = number of blades
Pitch ratios ranging from 0.6 to 1.04 have been tested. N = rotational speed of impeller, rev/s
For smaller pitch ratios, we do not expect the drag model Nf = rotational velocity of the fluid, rev/s
to hold for blade inclination angles smaller than 15 deg, Np = power number, P / p N 3 & , dimensionless
and the blades are almost tangent to the vector of the N, = rotational velocity of the fluid relative to the
oncoming fluid velocity. Then skin friction dominates on impeller, rev/s
both sides of the blades, resulting in an overall drag p = impeller pitch, m; or fluid pressure in Equation
larger than accounted for by the simple model of Equa- ( 5 ) , N/m2
tion (24). As expected, the power predicted for Zlokarnik's P = power consumed, W
system, in which case the impeller had a pitch ratio of r = radial position, m
0.5, is too low (see Table 3 ) . The power consumption Rb = radius of impeller blade, m
also increases considerably as the pitch ratio decreases R, = radius of the center of impeller blade, m
owing to the interrelation between the pitch and the Re = Reynolds number for mixing systems, d2NP/p,
blade length (see Table 1 ) . Finally, impellers with pitch dimensionless
ratios equal to 0.5 or smaller would have a low pumping Reb = Reynolds number about the blade, w2?rN,Rcp/p,
capacity or mixing rate, resulting in a very poor efficiency. dimensionless
tm = mixing time, s
Effect of the Blade Width V. = approaching fluid velocity, m/s
The blade width-to-diameter ratio varied from 0.09 to w = blade width, m
0.19 in this work. Doubling the blade width doubles
Greek Letters
the pumping capacity and the mixing rate. However, as
discussed, the power consumed is almost independent of 6
- = unit tensor
the blade width. The assumptions leading to the proposed = fluid viscosity, N.s/mZ
model become less valid as the ratio w / d is increased. p = fluid density, kg/m3
Good predictions are obtained for values of w / d 6 0.2, -
o = stress tensor, N/m2
but additional tests are required to confirm the model's Ji = blade inclination angle, deg
adequacy beyond this point. Although impellers with T = torque, N - m

AlChE Journal (Vol. 25, No. 3) May, 1979 Page 515

 LITERATURE CITED
Bourne, J. R., The Mixing of Powders, Pastes and Non-New-
tonian Fluids, Chem. Eng. (London), 180, CE202 ( 1964).
-, and H. Butler, Some Characteristics of Helical Im-
pellers in Viscous Liquids, AIChE Symposium Ser., 10, 89
(1965).
, Power Consumption of Helical Ribbon Im-
pellers in Viscous Liquids, Trans. Inst. Chem. Engrs., 47,
T263 ( 1969).
Carreau, P. J., I. Patterson, and C. Y. Yap, Mixing of Visco-
elastic Fluids with Helical Ribbon Agitators, I-Mixing Time
and Flow Patterns, Can. J. Chem. Eng., 54, 135 ( 1976).
Chavan, V. V., and J. I. Ulbrecht, Internal Circulation in
Vessels Agitated by Screw Impellers, Chem. Eng. J., 6,
213 (19734).
, Dynamics of Agitation of Non-Newtonian Liq-
uids by Close-Clearance Agitators, Ind. Eng. Chem. Proc-
ess Design Develop., 12, 472 ( 1973B).
Coyle, C. K., H. E. Hirschland, B. J. Michel, and J. Y. Oldshue,
Mixing in Viscous Liquids, AlChE J., 16, 903 (1970).
Danckwerts, D. V., Theory of Mixtures and Mixing, Appl.
Sci. Res., A3, 279 ( 1953).
Ford, L). E., R. A. Mashelkar, and J. Ulbrecht, Mixing Times
in Newtonian and Non-Newtonian Fluids, Proc. Tech. In-
ternational, 17, 803 ( 1972).
Gray, J. B., Batch Mixing of Viscous Liquids, Chem. Eng.
Progr., 59, No. 3, 55 ( 1963).
Hall, K. R., and J. C. Godfrey, Power Consumption by Helical
Ribbon Impellers, Trans. Inst. Chem. Engrs., 48, T20
(1970).
Holland, F. A., and F. S. Chapman, Liquid Mixing and Proc-
essing, Reinhold, New York (1966).
Hoogendoom, C. J., and A. P. Den Hartog, Model Studies
on Mixers in the Viscous Flow Region, Chem. Eng. Sci.,
22,1689 (1967).
Johnson, R. T., Batch Mixing of Viscous Liquids, Ind. Eng.
Chem. Process DeJign Develop., 6, 340 (1967).
Lacey, P. M. C., Developments in the Theory of Particle
Mixing, J. Appl. Chem. (London), 4, 257 (1954).
Moo-Young, M., K. Tichar, and F. A. L. Dullin, The Blending
Efficient-es of Some Impellers in Batch Mixing, AIChE J.,
18, 179 (1972).
Nagata, S . , M. Yanagimoto, and T. Yokoyama, Studies of the
Power Requirements of Mixing Impellers-111, Mem. Fac.
Eng., Kyoto U n h , 18,414 (1956).
, A Study of the Mixing of High-viscosity Liq-
uids, Chem. Eng. (Japan),21, 278 (1957).
Nagata, S., Mixing Principles and Applications,
.. Halstead, New
fork ( 1975). -
Oldshue, J. Y., Mixing-Annual Review, Ind. Eng. Chem.,
58. No. 11.50 1966 ).
Parker, N. H., Mixing, Chem. Eng., 71, No. 12, 165 (1964).
Reher, E., and R. Boehm, Ruehren Nicht-Newtonscher
Fluessigkeiten, Chem. Technik, No. 3, 136 ( 1970).
Rushton, J. H., How to Make Use of Recent Mixing Devel-
opments, Chem. Eng. Progr., 50, 587 (1954).
Ulrich, H., and H. Schreiber, Stirring in Viscous Liquids,
Chem. Ing. Tech., 39,218 (1967).
Voncken, R. M., Homogenization of Miscible Liquids-11,
Brit. Chem. Eng., 10, 12 and 179 (1965).
Whitaker, S., Introduction Fluid Mechanics, Prentice Hall,
Englewood Cliffs, N.J. ( 1968 ).
White, A. M., and E. Brenner, Studies in Agitation, Trans.
Am. InPt. Chem. Engrs., 30,585 (1934).
Yap, C. Y., Mixing of Viscoelastic Fluids with Helical-Ribbon
Agitators, Doctoral thesis, Ecole Polytechnique, Montreal,
Canada ( 1976).
Zlokamick, M., Suitability of Stirrers for the Homogenization
of Liquid Mixtures, Chem. Eng. Techn., 39, 539 (1967).
Manuscript received Ianmry 23, 1978; revision received January 2,
and accepted February 6, 1979.

Part 111. Non-Newtonian Fluids

C. Y. YAP
W. I A N PATTERSON
and
P. J. CARREAU
The use of helical ribbon agitators to mix viscous non-Newtonian fluids
has been investigated. A generalized model, based on an appropriate defini-
tion of effective viscosity, is proposed to predict power consumption. This
model is most successful with fluids that do not have a high degree of elas-
ticity.
It was found that the efficiency of mixing of pseudoplastic fluids was about
half of that of Newtonian fluids in the same mixer, while the efficiency of
mixing viscoelastic fluids was still lower and approximately independent of
the mixer geometry. Blade width was the primary variable affecting the mix-
ing efficiency on inelastic fluids.

SCOPE

The mixing of very viscous fluids can be efficiently ac-
complished with a helical ribbon agitator (HRA). A model
to predict the power consumption of the HRA mixing
Newtonian fluids has been reported by Patterson et al.
(1979); however, very viscous fluids are often non-New-
tonian or viscoelastic in nature. This paper reports the
Correspondence concerning this paper should be addressed to Pierre
Carreau. C. Y. Yap is with the University of Malaysia.
0001-1541-79-2553-0516-$00.85, 0 The American Institute of Chem-
ical Engineers, 1979.
experimental results of various HRAs mixing pseudo-
plastic and viscoelastic fluids. Six different mixer geome-
tries with three different fluids were examined. The fluids
varied from negligible elasticity (2% aqueous Natrosol)
to high elasticity (lvoaqueous Separan). The model pre-
viously derived has been generalized to include the mixing
of non-Newtonian fluids. This was accomplished by de-
fining an effective rate of deformation that is functionally
dependent on the agitator geometry rotational speed and

Page 516 May, 1979 AlChE Journal (Vol. 25, No. 3)